You are on page 1of 172

Vibration based damage detection in composite structures using computational intelligence tools

Obinna K. Ihesiulor

A thesis submitted in fulfillment of the requirements for the degree of Master of Engineering (Research)

Vibration based damage detection in composite structures using computational intelligence tools Obinna K. Ihesiulor A thesis

School of Engineering and Information Technology University College University of New South Wales Australian Defence Force Academy

31 August, 2012.

Originality Statement

’I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed:

Copyright and DAI Statement

’I hereby grant the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

  • I also authorise University Microfilms to use the 350 word abstract of

my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).

  • I have either used no substantial portions of copyright material in my

thesis or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.’

Signed:

Date:

Authenticity Statement

’I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are any minor variations in formatting, they are the result of the conversion to digital format.’

Signed:

Date:

i

Abstract

This research work deals with aspects concerned with delamination detection in com- posite structures as revealed by an approach based on vibration measurements. Variations in vibration characteristics generated in composite laminates indicate the existence of delamina- tions. This is because degradation due to delamination causes reduction in flexural stiffness and strength of a material and as a result, vibration parameters like natural frequency re- sponses are changed. Hence it is possible to monitor the variation in natural frequencies to identify the presence of delamination, and assess its size and location for online structural health monitoring (SHM). The approach in this thesis typically depends on undertaking the analysis of structural models implemented by finite element analysis (FEA). FE models also known as the simu- lator are used to compute the natural frequencies for delaminated and intact specimens of composite laminates. The FE models are validated using the analytical model. However, these FE models are computationally expensive and surrogate (approximation) models are introduced to curtail the computational expense. The simulator is employed to solve the inverse problem using different algorithms based on computational intelligence concepts. An artificial neural network model (ANN) is developed to solve the inverse problem for de- lamination detection directly and to provide surrogate models integrated with optimization algorithms (the gradient based local search and Real-coded Genetic Algorithm (RGA)). This approach is termed as surrogate assisted optimization (SAO) and it is seen that the engage- ment of surrogate models in lieu of the FE models in the optimization loop greatly enhances the accuracy of delamination detection results within an affordable computational cost. It also provides control when handling different variables. Meanwhile, to aid the building of effective surrogate models using substantial number of training datasets, K-means cluster- ing algorithm is harnessed and this effectively reduces the large training datasets usually required for ANN network training. Response surface methods (RSM) are also developed to directly solve the inverse problem. The principal advantage of the RSM is its ability to give physical mathematical models that are used to identify the size and location of delamination given any input changes in natural frequencies. A delamination detection strategy that uses K-means clustering algorithm for database selection and ANN, RSM and optimization algorithms integrated with surrogate models based on ANN have been successfully developed. It is demonstrated that these algorithms show immense potentialities for use in delamination damage detection scenarios when ap- plied to composite beams and plates. The algorithms successfully performed delamination detection given limited amount of training datasets. Prediction errors of the algorithms were quantified and they were shown to be robust in the presence of artificial errors and noise and

ii

even when applied to experimental and simulation data. Results clearly indicate remark- able accurate delamination damage detection capability of the algorithms. The algorithms in their inverse formulations are capable of predicting accurately delamination parameters. These algorithms should hence be employed for application in the domain of SHM where their small computational requirements could be exploited for online damage detection.

iii

Acknowledgments

My gratitude goes to all the individuals who made this thesis possible. First and foremost, I would like to thank my supervisor, Dr. Krishna Shankar for the invaluable opportunity given to me to be his student. I thank him immensely for believing in me and allowing me undertake research under his supervision. With a busy schedule and

many other research students to supervise, he has always been very responsive to my requests and always had great ideas for me and there has never been an occasion when I have knocked on his door and he did not give me time. My thanks also go to my joint-supervisor, Dr. Tapabrata Ray. Without his extraordinary optimization ideas and computational expertise, this thesis would have been a distant dream.

  • I express my gratitude to Zhifang Zhang for her codes used for ANSYS simulations. My

colleagues at Multi-Disciplinary Optimization (MDO) group who have enriched my research life in many ways, deserve a special mention. Thanks to Asaf for his great help in providing the basic understanding of L A T E X for writing this thesis.

  • I would also like to acknowledge the help and support from the staff members. Vera Berra, Craig Edward and Liz Carey’s administrative help, suggestion and kindness is highly

appreciated, as is the support services from the Library and ICTS department.

Special thanks and love to Mum and Dad, my siblings and other family members for their endless inspiration, love and support throughout my studies. Their support through these years are greatly appreciated.

  • I offer my regards to colleagues and friends @ UNSW@ADFA; Adura, Ahmed, Essam,

Moysen, Chigozie, Chengjun, Ram, Sayem, Vishal, and many others who supported and helped me throughout my Masters program. My interactions with them and others too numerous to mention have been very fruitful. My housemates at my place of residence have also played a positive role in my finishing well. I would also like to express my gratitude to my family friends in Canberra who have made my social and academic life balanced. Not left out, I acknowledge the financial support from UNSW@ADFA that made this venture worthwhile. It is impossible to remember all, and I apologize to those I have inad- vertently left out. Finally, thank you all and thank God for making my journey thus far an unforgettable one!!

iv

Publications arising from the thesis work

Journal papers

  • 1. Obinna K. Ihesiulor, Krishna Shankar, Zhifang Zhang, and Tapabrata Ray. “Delam- ination detection with noise polluted measurements of natural frequencies using computational intelligence concepts”. Computers and Structures (under review).

  • 2. Obinna K. Ihesiulor, Krishna Shankar, Zhifang Zhang, and Tapabrata Ray. “Valida- tion of algorithms for delamination detection in composite structures using experimental data”. Composite Structures (under review).

  • 3. Obinna K. Ihesiulor, Krishna Shankar, Zhifang Zhang, and Tapabrata Ray. “Efficien- cies of algorithms for vibration-based delamination detection: A compara- tive study”. Optimization and Engineering (under review). Conference Papers

  • 1. Obinna K. Ihesiulor, Krishna Shankar, Zhifang Zhang, and Tapabrata Ray. “Effec- tiveness of artificial neural networks and surrogate-assisted optimization techniques in delamination detection for structural health monitoring”. In Proceedings of the 23rd IASTED International Conference on Modelling and Simula- tion (MS 2012), July, 2012, Banff, Canada.

  • 2. Obinna K. Ihesiulor, Krishna Shankar, Zhifang Zhang, and Tapabrata Ray. “Delami- nation detection using methods of computational intelligence”. In Proceedings of the sixth Global Conference on Power Control and Optimization (PCO 2012), Las Vegas, USA, August, 2012.

  • 3. Zhifang Zhang, Obinna K. Ihesiulor, Krishna Shankar, and Tapabrata Ray. “Compar- ison of Inverse Algorithms for Delamination Detection in Composite Lam- inates”. In Proceedings of the ASME 2012 Conference on Smart Materials, Adaptive Structures and Intelligent Systems (SMASIS 2012), September, 2012, Stone Mountain, Georgia, USA. (Accepted and to be published).

Contents

Abstract

i

Acknowledgments

 

iii

Publications arising from the thesis work

 

iv

List of Figures

x

List of Tables

xiv

Nomenclature

xix

  • 1 Introduction

1

  • 1.1 Background

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

1

  • 1.2 Structural health monitoring (SHM)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

3

  • 1.3 Approach to thesis

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

5

  • 1.4 Limitations of current approaches

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

8

  • 1.5 Objectives of research

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

9

  • 1.6 Research scope

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

10

  • 1.7 Summary of contributions

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

11

  • 1.8 Organization of thesis

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

12

  • 2 Literature review

 

14

  • 2.1 Overview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

14

 

v

CONTENTS

vi

  • 2.2 Overview of SHM methods for damage detection

 

.

16

  • 2.3 Potential SHM methods except vibration based methods

 

.

.

.

.

.

.

.

.

.

.

.

18

  • 2.4 Methods based on vibration monitoring

 

.

19

  • 2.4.1 Studies on the effects of delaminations on natural frequencies

 

.

20

  • 2.4.2 Solution to the inverse problem

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

24

  • 2.4.2.1 Intelligent inverse algorithms for vibration based damage de- tection . . . . . . .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

24

 

Direct optimization based methods

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

24

Surrogate Assisted Optimization (SAO) based methods

 

.

26

Artificial-neural network based methods

.

27

  • 2.4.2.2 RSM as inverse method for vibration based damage detection

29

  • 2.5 Summary of research gaps

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

30

  • 3 Mathematical formulation and modeling of composite laminates for beams and plates

32

  • 3.1 Overview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

32

  • 3.2 Mathematical formulation of the optimization problem

.

.

.

.

.

.

.

.

.

.

.

.

33

  • 3.2.1 Solution methodology

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

35

  • 3.3 Modeling of the laminated composite beam

 

.

37

  • 3.3.1 Finite Element (FE) Analysis of the composite beam

 

.

.

.

.

.

.

.

.

.

37

  • 3.3.2 Theoretical modeling of the laminated composite beam

.

.

.

.

.

.

.

.

42

  • 3.3.3 Validation of the FE Analysis for the composite beam

 

.

45

  • 3.4 FE modeling of composite plates

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

46

  • 4 Algorithms for solution of the inverse problem

 

49

  • 4.1 Overview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

49

  • 4.2 Optimization search algorithms

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

49

  • 4.2.1 Gradient based Local Search (GBLS)

 

.

50

CONTENTS

vii

  • 4.2.2 Global optimizer based on evolutionary algorithm (EA)

.

51

  • 4.3 Database creation for ANN training using K-Means clustering algorithm

.

.

55

  • 4.3.1 Preparation of database used for ANN training

.

.

.

.

.

.

.

.

.

.

.

.

56

  • 4.4 Artificial Neural Network (ANN) as an inverse solver and surrogate creator .

57

  • 4.4.1 Introduction to Artificial Neural Network (ANN)

.

.

.

.

.

.

.

.

.

.

.

57

  • 4.4.2 Pros of ANN

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

59

  • 4.4.3 Cons of ANN

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

60

  • 4.4.4 Modifications to the basic ANN

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

61

  • 4.4.5 The steps in designing ANN model

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

62

  • 4.4.6 Generalization enhancement

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

64

  • 4.4.7 Trial training algorithms and learning functions

.

.

.

.

.

.

.

.

.

.

.

.

65

  • 4.4.8 Performance study between single and ensemble neural nets

.

.

.

.

.

65

  • 4.4.9 ANN configuration to analyze the best network architectures

.

.

.

.

.

66

  • 4.4.9.1 Training performance of trial neural networks for the forward problem

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

67

  • 4.4.9.2 Training Performance of trial neural networks for the inverse problem

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

69

  • 4.4.10 Study of the effect of transfer functions on network performance

.

70

  • 4.4.11 Data preprocessing

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

70

  • 4.4.12 Summary of selected best neural network architectures

.

.

.

.

.

.

.

.

71

  • 4.5 Surrogate approach

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

71

  • 4.5.1 Validation of the surrogate model

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

72

  • 4.6 Response Surface Methodology (RSM)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

74

  • 4.6.1 Introduction to RSM

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

74

  • 4.6.2 Advantages of RSM .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

75

  • 4.6.3 Analysis of a first-order model response surface

.

.

.

.

.

.

.

.

.

.

.

.

75

  • 4.6.4 Analysis of a second-order model response surface

.

76

CONTENTS

viii

  • 4.6.5 Validation and adequacy check of the developed models

 

.

78

  • 5 Results and discussion

 

81

  • 5.1 Overview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

81

  • 5.2 2-Variable problem–Prediction of delamination location and size at known

 

interface

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

81

  • 5.2.1 Application of ANN for delamination prediction

.

.

.

.

.

.

.

.

.

.

.

.

82

  • 5.2.2 Application of RSM for delamination prediction

.

.

.

.

.

.

.

.

.

.

.

.

83

  • 5.2.3 Optimization without surrogates

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

84

  • 5.2.4 Surrogate Assisted Optimization (SAO)

 

.

85

 
  • 5.2.4.1 Optimization using RGA with surrogates

 

.

86

  • 5.2.4.2 Optimization using gradient based local search (GBLS) method

 

with surrogates

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

88

  • 5.3 3-Variable problem–Prediction of delamination location and size at unknown

 

interface

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

89

  • 5.4 Validation with published experimental results

 

.

91

  • 5.4.1 Experimental validation with Okafor et al.’s results

 

.

.

.

.

.

.

.

.

.

.

91

  • 5.4.2 Experimental validation with Su et al.’s published results

 

.

91

  • 5.5 Effect of artificial errors and noise .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

93

  • 5.5.1 Effect of artificial errors

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

93

  • 5.5.2 Effect of artificial noise .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

95

  • 5.6 5-Variable problem–Delamination detection in composite plates

.

.

.

.

.

.

.

100

  • 5.6.1 Prediction of delamination parameters for the composite plate lami- nates using RGA with surrogates

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

100

  • 6 Comparison of delamination prediction efficiencies of different algorithms 102

  • 6.1 Overview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

102

  • 6.2 Algorithm 1 - ANN

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

104

CONTENTS

ix

 
  • 6.3 Algorithm 2 - RSM

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

105

  • 6.4 Algorithm 3 - RGA WoS .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

105

  • 6.5 Algorithm 4 - RGA WS

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

106

  • 6.6 Algorithm 5 - GBLS WoS

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

107

  • 6.7 Algorithm 6 - GBLS WS

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

108

  • 6.8 Summary of comparative results

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

109

7

Conclusions and recommendations

 

114

  • 7.1 Conclusions

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

114

  • 7.2 Recommendations for future work .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

118

References

 

120

A Trial performance of different ANN network architectures, transfer func- tions, training functions and scaled values

127

 

A.1

Overview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

127

B

Validation of the surrogate models

 

141

B.1

Overview

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

141

List of Figures

  • 1.1 Approach to thesis.

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

8

  • 2.1 Outline for literature

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

16

  • 3.1 Schematic of solution methodology.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

36

  • 3.2 Sample lay plot display for [0/90/0/90]s sequence of fiber orientation in the

 

laminate

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

39

  • 3.3 Locations of different interfaces of the [0/90/0/90]s composite laminates

 

.

40

  • 3.4 3D plots of frequency shifts (dF) as functions of delamination location (X)